Structural characterization of horseradish peroxidase using EXAFS

James E. Penner-Hahn,*'a*b Kim Smith Eble,lc,d Thomas J. McMurry,lb Mark Renner,le. Alan L. Balch,le John T. Groves,lb>' John H. Dawson,*lf and Keith ...
0 downloads 0 Views 1001KB Size
7819

J . Am. Chem. SOC.1986,108, 7819-7825

Structural Characterization of Horseradish Peroxidase Using EXAFS Spectroscopy. Evidence for Fe=O Ligation in Compounds I and I1 James E. Penner-Hahn,*'a*b Kim Smith Eble,lc,dThomas J. McMurry,lb Mark Renner,le Alan L. Balch,le John T. Groves,lb>'John H. Dawson,*lf and Keith 0. Hodgson*'* Contribution f r o m the Department of Chemistry, The University of Michigan, Ann Arbor, Michigan 48109- 1055, the Department of Chemistry, The University of South Carolina, Columbia, South Carolina 29208, the Department of Chemistry, University of California, Davis, California 9561 6 , and the Department of Chemistry, Stanford Unioersity, Stanford, California 94305. Received May 5, I985

Abstract: Extended X-ray absorption fine structure spectroscopy has been utilized to determine the structural environment of the heme iron sites in horseradish peroxidase compounds I and 11. For comparison, analogous studies have been undertaken on putative ferry1 (FeIV=O) porphyrin model compounds and on crystallographically characterized CrrV=O and CrV=N porphyrins. In a preliminary communication, we suggested that a short ca. 1.6 A Fe-0 bond is present in the high valent forms of both the enzyme and the synthetic porphyrins. The present work demonstrates unambiguously that a short, ca. 1.64 A, Fe-O bond length is present both in HRP compounds I and I1 and in their synthetic analogues. This structure is consistent only with an oxo-ferry1 (Fe=O) complex as the active oxygen species in horseradish peroxidase. The structural details, their implications for heme protein mediated oxygen activation, and the difference between our results and those recently published by other workers (Chance, B.; Powers, L.; Chinp. Y . :Poulos. T.; Schonbaum, G. R.; Yamazaki, I.; Paul, K. G. Arch. Biochem. Biophys. 1984, 235, 596-61 1) are discussed.

The transport and activation of dioxygen by heme proteins is a subject of continuing interest. For proteins which activate dioxygen, the prototype being cytochrome P-450, the general mechanism is believed to involve transient formation of an activated iron-oxo specie^.^^^ The crystal structure of ferric P-4.50 has recently appeared," however, the putative Fe(0) intermediate for this enzyme has not yet been isolated, and if isolated, it is unlikely to be sufficiently stable to permit crystallographic structure determination. However, the active oxygen species in P-450 is generally believed to be structurally similar to the high-valent intermediates formed during the catalytic cycle of the peroxidase enzyme^.^ As part of a continuing program to characterize the mechanism of heme-mediated oxygen a ~ t i v a t i o n , ~ , ~ we have undertaken an X-ray absorption spectroscopy (XAS)7 study of the high-valent intermediates formed during the catalytic cycle of horseradish peroxidase (HRP). Horseradish peroxidase is a heme-containing enzyme that catalyzes the oxidation of phenolic substrates, utilizing hydrogen peroxide as the ultimate electron acceptor. The resting form of H R P contains a high-spin ferric heme. During the catalytic cycle (1) (a) Stanford University. (b) University of Michigan. (c) University of South Carolina. (d) Present Address: Department of Biochemistry, Wake Forest University, Winston-Salem, N C 27103. (e) University of California. (f) Present Address: Department of Chemistry, Princeton University, Princeton, NJ 08540. (2) Malmstrom, B. G. Annu. Rev. Biochem. 1982,51,21-59. (3) Eble, K. S.; Dawson, J. H. Adu. Inorg. Bioinorg. Mech. 1986,4, 1-64. (4) Poulos, T. G.; Finzel, B. C.; Gunsalus, I. C.; Wagner, G. C.; Kraut, J. J. Biol. Chem. 1985,260, 16122-16130. (5) (a) Hahn, J. E.; Hodgson, K. 0.;Andersson, L. A.; Dawson, J. H. J. Biol.Chem. 1982,257,10934-10941. (b) Dawson, J. H.; Andersson, L. A,; Hodgson, K. 0.; Hahn, J. E. In Cytochrome P-450 Biochemistry, Biophysics, and Environmental Implications; Heitanen, E., Laitinen, M., Hanninen, 0.. Eds.; Elsevier: Amsterdam, 1982; pp 589-596. (c) Dawson, J. H.; Andersson, L. A,: Davis, I. M.; Hahn, J. E. In Biochemistry, Biophysics, and Regulation of Cytochrome P-450;Gustafsson, J. A,, Carlstead-Duke, J., Mode, A., Rafter, J., Eds.; Elsevier: Amsterdam, 1980; pp 565-572. (6) Penner-Hahn, J. E.; McMurry, T.; Renner, M.; Latos-Grazynski, L.; Eble, K. S.; Davis, I. M.; Balch, A.; Groves, J. T.; Dawson, J. H.; Hodgson, K. 0. J. Biol. Chem. 1983,258, 12761-12764. (7) Abbreviations: XAS, X-ray absorption spectroscopy; EXAFS, extended X-ray absorption fine structure; XANES, X-ray absorption near edge structure: EPR, electron paramagnetic resonance: ENDOR, electron-nuclear double resonance; HRP, horseradish peroxidase; TTP, dianion of 5,10,15,20-tetra-m-tolylporphyrin; TPP, dianion of 5,10,15,20-tetraphenylporphyrin; TMP, dianion of 5,10,15,20-tetramesitylporphyrin.

0002-7863/86/ 1508-78 19$01.50/0

Scheme I

HRP-IIe

HRP-I

AH*

(see Scheme I), H R P is initially oxidized by two electrons to a green species known as compound I (HRP-I). One-electron reduction of HRP-I yields the red species known as compound I1 (HRP-11), which retains one oxidizing equivalent above the resting enzyme. The intermediates in the H R P reaction cycle have been subjected to extensive physical and chemical characterization.&I0 The resting enzyme contains a five-coordinate heme, with the fifth coordination site occupied by an axial imidazole ligand.4*"-13 HRP-I is two oxidation equivalents above ferric iron and therefore formally contains FeV. Magnetic susceptibility measurements14 indicate that there are three unpaired electrons (S = 3/2) in HRP-I. This has been explained by ferromagnetic coupling of low-spin Fe" (S = 1) to a porphyrin T-cation radical (S = Consistent with this assignment is the similarity of the optical spectrum of HRP-I and those of synthetic porphyrin ~ - c a t i o n s . I ~ Investigations of HRP-I by NMRI6 and ENDORI7 as well as iterative extended Huckel (IEH) calculation^'^ all favor the T (8) Dunford, H. B.; Stillman, J. S. Coord. Chem. Rev. 1976,19, 187-251. (9) Hewson, W. D.; Hager, L. P. In The Porphyrins: Dolphin, D., Ed.; Academic: New York, 1979; Vol. 7, pp 295-332. (10) Frew, J. E.: Jones, P. Adv. Inorg. Bioinorg. Mech. 1984, 3, 175. (1 1) Yonetani, T.; Yamamoto, H.; Erman, J. E.; Leigh, J. S.; Reed, G. H. J . Biol. Chem. 1971,247,2447-2455. (12) Mauk, M. R.; Girotii, A. W. Biochemistry 1974, 13, 1757-1763. (13) Kobayashi, N.; Nozawa, T.; Hatano, M. Biochim. Biophys. Acta 1977,493, 340-351. (14) Theorell, H.; Ehrenberg, A. Arch. Biochem. Biophys. 1952, 41, 442-46 1. (15) Dolphin D.; Forman, A.; Borg, D. C.; Fajer, J.; Felton, R. H. Proc. Natl. Acad. Sci., U.S.A. 1971,68,614-618. (16) LaMar, G.; de Ropp, J. S.: Smith, K. M.; Langry, K. C. J. B i d . Chem. 1981,256,237-243. (17) Roberts, J. E.; Hoffman, B. M.; Rutter, R.; Hager, L. P. J. B i d . Chem. 1981,256,2118-2121.

0 1986 American Chemical Society

Penner-Hahn et ai.

1820 J . Am. Chem. SOC.,Vol. 108, No. 24, 1986

that HRP-I and HRP-11, and related compound I and compound I1 models, had identical structures from the perspective of XAS. At that time, we reported only preliminary results for the EXAFS analysis. Recently, Chance et al. have reported the results of an XAS study of HRP-I, HRP-11, and related catalase complexes.32 While 0 N confirming our results for HRP-I, this work suggested that HRP-I1 I 2 has a significantly different structure, characterized by a much longer (1.93 A) Fe-0 interaction. On the basis of this metric information, Chance et al. proposed a peroxidase mechanism N 0 involving formation of a double bonded (Fe=O) complex in HRP-I and, on reduction, a single bonded Fe-0 complex in HRP-11. This mechanism is in conflict with findings from our N N GrN laboratories6 and with other evidence21*22 that the Fe sites in HRP-I 4 and HRP-I1 are very similar. In addition, this new mechanism, -5 which would indicate hydroxyl ligation in HRP-XI, contradicts a substantial body of e ~ i d e n c e ~ ~favoring - ~ ' * ~ ' oxo ligation in this cation radical formulation. HRP-I1 is formed by one-electron reduction of HRP-I (Scheme I) and is therefore formally Fe1V.19*20 complex. The Mossbauer spectra of HRP-I and HRP-I1 are very similar We report here the complete results of our EXAFS investigations of HRP-I and HRP-11. To improve the accuracy of our and are different from that of the ferric enzyme,21*22 supporting the assignment of Fe" for both of the oxidized compounds. analysis, we have used K2Fe04 as a model for Fe=O EXAFS. Since the Fe-0 distance in K2Fe04is known to be quite short Isotopic labeling studies have shown unequivocally that HRP-I (1.67 this compound is a good model for determining the and HRP-I1 contain one atom of o ~ y g e n ,but ~ ~the . ~protonation ~ state of this oxygen has been a matter of controversy. HRP-I is EXAFS parameters which characterize short Fe-0 bonds, such as those which we have found in HRP-I and HRP-11. In order generally believed to contain an oxo (double bonded oxygen) ligand, an assignment which is supported by ''0 ENDOR meato test the validity of our data analysis procedure, we have obtained s u r e m e n t ~ . Both ~ ~ oxo and hydroxyl ligation have been proposed and analyzed XAS data for a crystallographically characterized for HRP-11; however, recent NMR26 and resonance Raman2' high-valent oxo-chromium porphyrin that is thought to have a studies appear to be consistent only with the oxo assignment. structure analogous to those of HRP-I and HRP-11. These results The site of oxidation for metalloporphyrins (Le., metal centered support our earlier6 interpretation of the HRP-I and HRP-I1 vs. porphyrin centered) is quite sensitive to axial l i g a t i ~ n . ~ ~ *structures ~~ and stand in contrast with those of Chance et a1.j2 Thus, the presence of FeIv rather than an Fe"' porphyrin *-cation Experimental Section in HRP-11 indicates a strong ligand field. Although nitrido and Sample Preparation and Handling. Commercial horseradish percarbido ligands28and dialkoxy ligation30are able to stabilize FeIV oxidase (Type VI, Sigma) was further purfied by using the methods of with respect to an Fe"' porphyrin *-cation, I E H calculation^^^ Shannon et to A403/A275 greater than 3.56.35 Highly concentrated suggest that an oxo ligand, but not a hydroxy ligand, is able to (>5 mM) samples of HRP-I and HRP-I1 were generated* at ca. 0 "C in 5 m M potassium phosphate buffer (pH 7.0). Samples were prepared stabilize the FeIVconfiguration when the distal ligand is imidazole. in the EXAFS sample cells (2 X 4 X 25 mm Mylar36windowed L ~ c i t e ~ ~ A straightforward method to characterize the axial ligation in frames) immediately prior to XAS data collection. HRP-I was prepared HRP-I and HRP-11 is via the structure of the Fe site. The by addition of ethyl hydroperoxide to ferric HRP, giving a characteristic instability of these species precludes X-ray diffraction studies; green solution. Addition of 1 equiv of reductant (ascorbate) to HRP-I however, XAS is an ideal tool to investigate these structures. We resulted in formation of a red solution, characteristic of HRP-11. Sample have previously communicated6 X-ray absorption near edge composition was verified with UV-vis absorption spectroscopy at 4 " C structure and preliminary extended X-ray absorption fine structure on a diluted aliquot. By the UV-vis criterion, sample homogeneity was (EXAFS) results for H R P compounds I and I1 and related greater than 95% for both HRP-I and HRP-11. After preparation, the high-valent iron porphyrin model compounds. The first coordisamples were frozen and maintained at -80 OC throughout data collection. Following completion of XAS data collection, the protein samples nation environments of the Fe in these models (1-3) are illustrated were immediately reanalyzed by UV-vis absorption spectroscopy; no in Scheme 11. This work showed that both H R P species contain significant degradation in sample homogeneity was observed. highly oxidized iron porphyrins (>Fe"') and a coordination enHigh-valent iron porphyrin model compounds with formal oxidation vironment that can only be modeled by inclusion of a short (- 1.6 states Fe" and Fev (see Scheme 11) were prepared according to literature %.) Fe-0 interaction. A key feature of this work was the finding procedures. Complex 2, (N-MeIm)(TTP)FelV(0),is a formally Fe"

Scheme I1

n

,N-ul.l-N

(18) Hanson, L. K.; Chang, C. K.; Davis, M. S.; Fajer, J. J . Am. Chem. Soc. 1981, 103, 663-670. (19) Hewson, W. D.; Hager, L. P. J . Biol. Chem. 1979,254, 3182-3186. (20) Havashi. Y.:Yamazaki. I. J . Biol. Chem. 1979. 254, 9101-9106. (21) Schhz, C. E.; Devaney, P. W.; Winkler, H.; Debrunner, P. G.; Doan, N.; Chiang, R.; Rutter, R.; Hager, L. P. FEBS Lett. 1979, 103, 102-105. (22) Moss, T. H.; Ehrenberg, A.; Bearden, A. J. Biochemistry 1969, 8, 4159-41 62. (23) Schonbaum, G. R.; Lo, S. J . Biol. Chem. 1972, 247, 3353-3360. (24) Adediran, S. A.; Dunford, H. B. Eur. J . Biochem. 1983, 132, 147-150. (25) Roberts, J. E.; Hoffman, B. M.; Rutter, R.; Hager, L. P. J . Am. Chem. SOC.1981, 103, 7654-7656. (26) LaMar. G. N.: de Rooo. J. S.: Latos-Grazvnski. L.: Balch. A. L.: Johnson, R. B.; Smith, K. M.; Pa'rish, D.' W.; Cheng, R.-J. >. km. Chem. Soc. 1983, 105, 782-787. (27) Terner, J.; Sitter, A. J.; Reczek, C. M. Biochim. Biophys. Acta 1985, 828, 73-80. (28) English, D. R.; Hendrickson,D. N.; Suslick, K. S. Inorg. Chem. 1983, 22, 368-370. (29) English, D. R.; Hendrickson,D. N.; Suslick, K. S . Inorg. Chem. 1985, 24, 122-127. (30) (a) Groves, J. T.; Quinn, R.; McMurry, T. J.; Nakamura, M.; Lang, G.; Boso, B. J . Am. Chem. SOC.1985, 107, 354-360. (b) Groves, J. T.; Gilbert, J. A. Inorg. Chem. 1986, 25, 125-127. c31) Hanson, L. K., personal communication.

model for the Fe site in HRP-11. It was prepared as previously deby cleavage of a pperoxo porphyrin dimer. Complex 2 has a magnetic moment of 2.8 p . and ~ its ~ e ~l e ~ t r o n i c ',H ~ ~NMR,38.39and Mossbauer spectram all show close similarities to the corresponding spectra of HRP-11. All of this evidence is consistent with a six-coordinate FeIV porphyrin having axial imidazole and oxo ligation. The formally FeV models were prepared by oxidation of ferric porphyrins, with the two-electron oxidation of (TMP)Fe"' giving two dif(32) Chance, B.; Powers, L.; Ching, Y.; Poulos, T.; Schonbaum, G . R.; Yamazaki, 1.; Paul, K. G.Arch. Biochem. Biophys. 1984, 235, 596-61 1. (33) McGinnety, J. A. Acta Crystallogr. 1972, 828, 2845-2852. (34) Shannon, L. M.; Kay, E.; Lew, J. Y. J . B i d . Chem. 1966, 241, 21 66-2 172. (35) Eble, K. S., Ph.D. Thesis, University of South Carolina, 1985. (36) Mylar and Lucite are registered trademarks of E. I. du Pont de Nemours & Co., Inc. (37) Chin, D. H.; LaMar, G. N.; Balch, A. L. J . Am. Chem. Soc. 1980, 102, 4344-4350. (38) Chin, D. H.; Balch, A. L.; LaMar, G . N. J . Am. Chem. SOC.1980, 102, 1446-1448. (39) LaMar, G. N.; de Ropp, J. S.; Latos-Grazynski, L.; Balch, A. L.; Johnson, R. B.; Smith, K. M.; Parish, D. W.; Cheng, R.-J. J . Am. Chem. Soc. 1983, 105, 782-787. (40) Simmoneaux, G.; Scholz, W. F.; Reed, C. A,; Lang, G. Biochim. Biophys. Acta 1982, 716, 1-7.

Structural Characterization of Horseradish Peroxidase

J . Am. Chem. Soc., Vol. 108, No. 24, 1986

7821

-~

n

c

Y)

Q W X

L 2.0

4.0

6.0

k

8.0

10.0

12.0

(A-1 )

Figure 1. EXAFS spectra (left) and Fourier transforms (right) for HRP-I and HRP-11. EXAFS data are weighted by k3 to enhance the amplitude of high k oscillations. EXAFS and Fourier transform calculated as described in text.

ferent high-valent model compounds!' In acidic solution, oxidation with m-chloroperoxybenzoic acid gave a green, HRP-I model (1) formulated as L(TMP+)Fe'V0.41bs42 Under our conditions, L = methanol. The temperature and magnetic field dependence of the Mossbauer spectrum of 1 is indicative of an S = 1 FeIV that is strongly and isotropically coupled to an S = porphyrin Treatment of 1 with base, or oxidation of (TMP)Fe"' with iodosylbenzene, gave a different highvalent complex with a red color (3). On the basis of N M R and Mossbauer e ~ i d e n c e , '3~ is formulated as a bismethoxy complex of FerVTMP. As a complement to the Fe studies, the crystallographically characterized4345high-valent C r porphyrins (TTP)CrrVO(4) and (TTP)CrVN (5) were also studied. The high-valent Fe compounds (1-3)were examined as frozen toluene solutions while K2Fe04 and the Cr models (4, 5, and K2Cr04) were prepared as powders diluted with BN and pressed into a thin pellet supported by mylar tape. The thermal lability of compounds 1-3 (rapid decomposition for temperatures above -80 "C) necessitated the development of new sample handling techniques. As described p r e v i ~ u s l y , ~ samples were prepared in polyethylene cells which permitted XAS and N M R spectra to be measured without the need for thawing and transferring the samples. Data Collection and Reduction. X-ray absorption data were collected at the Stanford Synchrotron Radiation Laboratory on the wiggler beam line VII-3 and the focused beam line 11-2, under dedicated running conditions (3-3.5 GeV, 4 0 mA, 8-pole wiggler field at 13 kG). The radiation was monochromatized with a Si(220) double-crystal monochromator. The data for the chromium compounds (4,5, and K2Cr04) were measured in transmission mode with nitrogen-filled ion chambers (2.0-cm gap, 300 V) to monitor incident and transmitted flux. The remaining data were collected as fluorescence excitation spectra with an array of Mn-filtered NaI(TI) scintillation detectors to monitor the Fe K a emission.& The data reported are an average of between 4 (for models) and 15 (for proteins) scans, comprising irradiation times of about 2-8 h/sample. Energy calibration was by the internal calibration technique:' using Fe and Cr foils, for Fe and Cr XAS respectively. Inflection points of the metal foil calibrations were assigned as 71 11.3 eV (Fe) and 5988.8 eV (Cr). The averaged data were reduced by subtracting a smooth polynomial pre-edge, extrapolated from a measured pre-edge; subtracting a two- or three-region cubic spline; and normalizing to the atomic falloff. The spline points were chosen empirically to minimize the residual low-frequency background while not reducing the observed amplitude of the EXAFS. The spline and its first and second derivatives were examined visually to ensure that the spline was not "fitting" the oscillatory com~~

(41) (a) Groves, J. T.; Haushalter, R. C.; Nakamura, M.; Nemo, T. E.; Evans, B. J. J . Am. Chem. Soc. 1981, 103, 2884-2886. (b) Groves, J. T.; McMurry, T. J. Reo. Port. Quim. 1985, 27, 102-103. (42) Boso, B.; Lang, G.; McMurry, T. J.; Groves, J. R. J . Chem. Phys. 1983, 79, 1122-1126. (43) Groves, J. T.; Kruper, W. J., Jr.; Haushalter, R. C.; Butler, W. M. Inorg. Chem. 1982, 21, 1363-1368. (44) Budge, J. R.; Gatehouse, B. M.K.; Nesbit, M. C.; West, B. 0. J . Chem. SOC.,Chem. Commun. 1981, 370-371. (45) Groves, J. T.; Takahashi, T.; Butler, W. M. Inorg. Chem. 1983, 22, 884-887. (46) Cramer, W. P.; Scott, R. A. Reo. Sci. Instrum. 1981, 52, 395-399. (47) Scott, R. A.; Hahn, J. E.; Doniach, S.;Freeman, H.C.; Hodgson, K. 0. J . Am. Chem. SOC.1982, 104, 5364-5369.

2.0

4.0

6.0

k

8.0

(a-1)

10.0

12.0

0.0

1.0

2.0

R

3.0

(a)

4.0

5.0

Figure 2. EXAFS spectra (left) and Fourier transforms (right) for 1, 2, and 3. EXAFS data are weighted by k3 to enhance the amplitude of high k oscillations. EXAFS and Fourier transform calculated as described in text. The EXAFS spectra show pronounced sensitivity to axial ligation, while the Fourier transforms show only a shift to lower R and a slight increase in peak width between 1 or 2 and 3 (see text).

i 2.0

~

4.0

6.0

k

8.0

(a-I )

10.0

12.0

0.0

1.0

2.0

R

3.0

(a)

4.0

5.0

Figure 3. EXAFS spectra (left) and Fourier transforms (right) for 4 and 5. EXAFS data are weighted by k 3 to enhance the amplitude of high k oscillations. EXAFS and Fourier transform calculated as described in text. The two principal peaks in the Fourier transform at R 1.5 8, represent Cr=O (Cr=N) and Cr-Npyrroler respectively.

-

ponent of the absorption. The normalized, background subtracted data were converted from energy E to photoelectron wave-vector k = [2m(E - E 0 / h 2 ] ' /with * E, values of 7130 eV (Fe) and 6010 eV (Cr). Fourier transforms of the EXAFS data were calculated by numerical integration with k3 weighted data. The EXAFS data and the corresponding Fourier transforms for HRP-I and HRP-11, for 1-3, and for 4 and 5 are shown in Figures 1-3, respectively.

Data Analysis The observed EXAFS x ( k ) c a n b e expressed a s

w h e r e N , is t h e n u m b e r of scatterers in t h e i t h shell, F, is t h e photoelectron backscattering amplitude of t h e ith shell, u2 is t h e mean-square variation in t h e absorber-scatterer distance, R,,, X is t h e mean-free p a t h length for t h e photoelectron, a n d 4 a s ( k )is t h e n e t phase shift i n t h e photoelectron wave during scattering. The sum is taken over all shells of scatterers, where a shell consists of some n u m b e r of indistinguishable a t o m s a t t h e s a m e (or indistinguishable) distance from t h e absorber. Typically, first shell bond lengths a n d coordination n u m b e r s c a n be determined t o f0.02A a n d *20%, r e s p e c t i ~ e l y . I~n~practice, ~~~ differences of one o r two in atomic number cannot be distinguished via E X A F S , t h u s a shell described below a s nitrogen (N) could b e d u e t o C, N, o r 0. For two shells of t h e s a m e atomic type, t h e approximate n c e k,,, for resolving power of EXAFS is AR 1 ~ / 2 k , , , . S~i ~ (48) Cramer, S . P.; Hodgson, K. 0. frog. Inorg. Chem. 1979, 25, 1-39. (49) Lee,P. A,; Citrin, P. H.; Eisenberger, P.; Kincaid, B. M. Reo. Mod. f h y s . 1981, 53, 769-806.

7 8 2 2 J. Am. Chem. SOC.,Vol. 108, No. 24, 1986 EXAFS data from dilute samples in typically 13-15 A-1, this places a lower limit of -0.1 A on the resolvable distance between shells; in practice, noise in the data results in an effective resolution closer to 0.2 A. Two different approaches were taken in the analysis of the EXAFS data. Both approaches involve determining the amplitude, A,,(k), and phase, +,,(k), which are expected for a particular absorber-scatterer interaction. The function A,,(k) incorporates all of the factors which affect the amplitude, except N, and Ras. As shown in eq 1, A,,(k) depends on the identity of the absorber and the scatterers and on the length and strength of the absorber-scatterer bond. The phase shift of the photoelectron wave, ,+, depends only on the identity of the absorber and the scatterer. Once determined, the functions A,,(k) and +,,(k) are used in conjunction with a nonlinear curve-fitting algorithm to model the observed EXAFS for an unknown compound. This approach makes use of the fact that amplitude and phase functions are reasonably transferrable form compound to compo~nd,"~ provided that the chemical environments are similar. It is possible to calculate a b initio the expected A,,(k) and +,,(k);SOhowever, the calculated values generally require adjustment in order to correctly model experimental EXAFS.49 This adjustment frequently takes the form of a linear scale factor, although more complicated functions are necessary to accurately reproduce the experimental data. In both of our analysis procedures,we have used experimental EXAFS of crystallographically characterized model compounds to empirically or semiempirically determine A,,(k) and +,,(k). In the first method, a six-parameter function was fitted to the measured EXAFS for model compounds.48 The models used were Fe"(TPP) and Fe"'(TPP)(imida~ole)~for Fe-N and K2Fe04for Fe-O. This parametrized function was then used as a reference when fitting the EXAFS of unknown samples. When fitting unknowns, the number of scatterers and the absorber-scatterer distance were adjusted as variable parameters. The principal advantage of this approach is the small number of variable parameters per shell and the corresponding decrease in parameter correlation problems. Compared to the two variable parameters in the empirical approach,"8 some ab initio methods use as many as four variable parameters per shell5' (scale factor, N,, R,,, and uas). For multishell compounds (e.g., HRP-I) this can lead to parameter correlation problems. The primary difficulty with our empirical procedure is its limited flexibility, both in accurately reproducing the empirical A,,(k) and 4 , , ( k ) functions and in varying structural parameters when fitting unknowns. The second analysis procedure was combination of empirical and semiempirical methods using a tabulated rather than a parametrized expression for the EXAFS amplitude and phase functions. These tabulated values permit more accurate reproduction of the details of the amplitude and phase functions. In the case of Fe-N EXAFS, ab initio amplitudes and phasesm were used to fit the experimental Fe-N EXAFS from FelI(TPP). This fit gave the optimum values for a linear scale factor (0.34) and AE0(-12 eV). In a procedure similar to Teo's FABM method,52 the empirically optimized scale factor and Eo were held fixed for all subsequent fitting. The values for the scale factor and AEo which were determined for Fe-N proved to be transferable to fitting of Cr-N EXAFS. When fitting unknown compounds, N, was set at a chemically reasonable value (i.e., 4 or 5) and R,, and baswere optimized. Since u,, is not known for our model compounds, the u, values determined in this way represent only the difference in a,, between the model and the unknown. This procedure was attempted for M - 0 EXAFS, using experimental EXAFS from K2Cr04and K2Fe04.53 In this case, (50) Teo, B.-K.; Lee, P. A. J . Am. Chem. SOC.1979, 101, 2815-2832. (5 1) See, for example: Antonio, M. R.; Teo, B.-K.;Cleland, W. E.; Averill, B. A. J . A m . Chem. SOC.1983, 105, 3477-3484. (52) Teo, B. K.; Antonio, M. R.; Averill, B. A. J . Am. Chem. SOC.1983, 105, 3751-3162. (53) Structural parameters for K2Cr04and K2Fe04and from Hoppe et al. (Hoppe, M. L.; Schlemper, E. 0.;Murmann, R. K. Acru Crystallop. 1982, B38, 2237-2239) and from ref 33, respectively.

Penner-Hahn et al. 5.00,

1

I

I .oo

1

t 0.001 3.0

I '

1

5.0

"

7.0

"

9.0

"

11.0

'

i

13.0

R(A) Figure 4. Empirical (heavy line) and theoretical (light line) amplitude functions F,(k) for Fe-0 (see eq 1 for definition). Empirical F,(k) determined from complex back-transform of filtered first shell peak in Fourier transform of K2Fe0, EXAFS data. Theoretical function from ref 50. Theoretical function has been multiplied by the best fit scale factor (see text) of 0.30 to facilitate comparison. Amplitude functions multiplied by kZ to approximately correct for k dependence of backscattering amplitude.

however, it was not possible to obtain a good fit of the experimental data with the ab initio functions. The best fit was obtained for AEo = -20 eV, scale factor = 0.30, and u, = 0.0 A. The reasons for the poor quality of this fit are apparent from a comparison of the empirical and theoretical values of F , ( k ) . As shown in Figure 4, these functions have dramatically different shapes, which cannot be brought into agreement by means of a simple exponential factor. The reasons for this difference are unclear, although they may be related to the high oxidation state of the Fe and the strong bonding between the Fe and the 0. The data in Figure 4 indicate that, when fitting data for compounds containing Fe=O bonds, the experimentally determined Fe-0 amplitude function will probably give a more accurate fit than could be obtained with empirically adjusted ab initio parameters. The A,(k) and $,,(k) functions for M - 0 EXAFS were determined from the experimental model compound EXAFS data with a Fourier backt r a n ~ f o r m a t i o n . ~These ~ functions were then used in the same manner as the a b initio functions, with AEo fixed at zero and the scale factor fixed at 1. As before, R,, and Au,, were optimized when fitting unknowns.

Results As we have noted previously, the XANES spectra for HRP-I, HRP-XI, 1, and 2 are all extremely similar. In each case, the principal absorption discontinuity occurs at a higher energy than that found for Fe"' porphyrins. This suggests6 the presence of FelV in all four species. The spectra for HRP-I and HRP-I1 are superimposable within the noise level of the data, and those for 1 and 2 are quite similar.6 These spectra all lack the intense pre-edge feature frequently associated with metal-oxo porphyrins; however, recent calculationsSSconfirm that the XANES spectra for HRP-I, HRP-11, 1, and 2 are consistent with an FeIV=O(porphyrin) structure. The similarity of these XANES spectra demonstrates that the four different Fe sites must have similar structures. This interpretation is supported by the previously noted6 similarity in the EXAFS spectra of these compounds (Figure 1 and 2). The present work extends these qualitative observations to include the crystallographically characterized high-valent Cr porphyrins 4 and 5 (see Scheme I1 for structures). As shown in Figures 1-3, the EXAFS spectra for HRP-I, HRP-11, 1, 2, 4, and 5 are all quite similar, with a characteristic principal maximum 6.1 A-I. The EXAFS spectrum for 3, with a maximum at k

-

(54) Eisenberger, P.; Shulman, R. G.; Kincaid, B. M.; Brown, G. S.; Ogawa, S . Nature (London) 1978, 274, 30-34. (55) Penner-Hahn, J. E.; Benfatto, M.; Hedman, B.; Takahashi, T.; Doniach, S.;Groves, J. T.; Hodgson, K. 0.Inorg. Chem. 1986, 25, 2255-2259.

J . Am. Chem. SOC.,Vol. 108, No. 24, 1986 7823

Structural Characterization of Horseradish Peroxidase Table I. Iron-Porphyrin Structural Parameters’ method I sample M-NC M-0‘ 1 2.04 1.62-1.65 2 2.02 1.64-1.65 3 I .92 4 2.01 1.51 5 2.02 1.49 HRP-I 1.99 1.61-1.62 HRP-I1 1.99 1.60 6

method I1 M-NC M-0‘ 2.02 2.03 I .92 2.01

1.63-1.66 1.65-1.66

d

d

2.00 2.00

1.62-1.64 1.62-1.64

1.52

crystallograPhYb M-N‘ M-0‘

2.03 2.04

‘4-1.

-

6.4 %.-I,

I

w

n 3

1.57 1.57

I K 0 LL

2.005

1.604

Metal to nearest-neighbor bond lengths from analysis of EXAFS data except as indicated. EXAFS results from nonlinear curve fitting methods described in text. M-N is the average M-Npyrrolebond length. For EXAFS results, the sixth ligand (methanol or imidazole) is included in this average. M - 0 is the M=O bond length or, for 5, the C r G N bond length. bCrystallographic results from ref 32, 34, and 60. ‘M = Cr for 4 and 5, Fe in all other cases. “Not determined.

at k

I\

is noticeably different over the range k

-

5-9

Despite these differences, all of the Fe EXAFS spectra give fairly similar Fourier transforms (Figures 1 and 2). The main peak in the transform for 3 is slightly broader and is shifted to slightly lower R relative to those for 1 and 2; however, these differences are minor compared to the differences in the corresponding EXAFS spectra. This result illustrates the dangers of basing structural analyses on the simple qualitative appearance of the Fourier transform magnitudes. On the basis of a curvefitting analysis (vide infra) the Fe porphyrins are found to contain two shells of ligands; however, these shells are not resolved in the Fourier transforms. In contrast, the Fourier transforms for the Cr porphyrins (Figure 3) show the presence of two distinct shells of nearest-neighbors. This is not surprising, since the Cr porphyrins are five-coordinate and thus are expected to have shorter Cr-(axial ligand) bond lengths and longer Cr-Nmmle bond lengths than will be found in the six-coordinate Fe porphyrins. The curve-fitting analysis (vide infra) suggests that the inter-shell separation increases by 0.1 A between the Fe and the Cr porphyrins. Evidently, this increase in separation is adequate to permit resolution of the two different shells in 4 and 5. Note that despite the very different appearance of their Fourier transforms, the EXAFS spectra and structural parameters (vide infra) of 1, 2, 4, and 5 are quite similar. This again illustrates that Fourier transforms cannot reliably be used to infer structural details. The qualitative similarity in the EXAFS spectra for HRP-I, HRP-11, 1, 2, 4, and 5 suggests that the structures of the metal sites in all of these species must be very similar. In order to test this interpretation, we have carried out a complete curve-fitting analysis of the EXAFS data. All of the results discussed below are based on data for which a Fourier filter has been used to isolate the EXAFS due to only those atoms in the first coordination sphere. This procedure greatly simplifies the analysis and in our hands gives no loss of accuracy. As a check, all final fits were repeated with unfiltered data. No change was observed in the structural results; however, the rate convergence for these fits was significantly slower. The quality of the fits was assessed by calculating the root-mean-square deviation between the data and the fit. Chromium Model Compounds. Since they are structurally ~haracterized,~ the ~ -high-valent ~~ C r compounds are useful for testing the accuracy of our data analysis procedures. Attempts to fit the first-shell EXAFS for 4 or 5 with a single shell of N atoms at -2 A gave poor agreement with the experimental data. The mean-square deviation between data and fit decreased by a factor of 2 on inclusion of a shell of 0 atoms at a short distance (1.5-1.6 A). The final two shell fitting results for 4 and 5 are given in Table I. Iron Complexes. The optimum paarameters for Fe-N and Fe-0 were determined as described above. Fits were attempted with either one (N) or two (N 0) shells and with either four or five nitrogens in the nitrogen shell. The distances determined

(0

z a L

R (8) Figure 5. Fourier transforms of best fits to HRP-I data. Fits calculated with method I. Mean-square derivation, D, between data and fit given by D = (C[(x,?,-x f i , ) k ’ ] * / ( N -l))1/2. (A) HRP-I data (dark) and best fit (light) obtained when only a single shell of nitrogens (6.3 N at 1.99 A, D = 1.2) was used. (B) HRP-I data (dark) and best fit (light) obtained when only a single shell of oxygen (1.9 0 at 1.61 A,D = 2.8) was used. (C) HRP-I dark (dark) and best fit (light) obtained when both a shell of nitrogen and a shell of oxygen (5.6 N at 1.99 A;1.1 0 at 1.61 A; D = 0.5) were used. (D) Best fit from two-shell fit (dark) and contributions from N and 0 shells. Note the absence of a resolvable Fe-0 peak in the simulated data for two shells separated by 0.4 A.

for these different fits were all very similar. The range of values reported in Table I represents the extreme values obtained for a particular distance in any of the different fits which were performed. The uasvalues determined in this way were quite sensitive to the details of the fit and in particular were strongly correlated to the number of scatterers used in the fit. As noted above, the high-valent models 1 and 2 have essentially indistingusishable EXAFS spectra. These data could not be adequately fit with only a single shell of nitrogens. However, inclusion of a second shell of oxygen at 1.6 A decreased the mean-square deviation between data and the fit by a factor of 2 to 4. The results of these two-shell fits are given in Table I. Essentially identical results were observed for HRP-I and HRP-11. These samples have nearly superimposable EXAFS spectra which cannot be fitted adequately with a single shell of atoms. In both cases, using either fitting method, the improvement in the mean-square deviation on inclusion of a second shell was at least a factor of 2. A typical set of fits, for HRP-I with method I, is shown in Figure 5. These fits demonstrate graphically the requirement of two shells in order to obtain a reasonable fit. As shown in Figure 5, the Fourier transforms for this range of data do not give resolvable peaks for the N at 2.0 A and the 0 at 1.6 A. Nevertheless, these shells can be resolved by curve-fitting methods. The quantitative curve-fitting results for HRP-I and HRP-I1 (Table I) are very similar to those obtained for their putative models 1 and 2. This is consistent with the close similarity between the EXAFS spectra of the protein and those of the synthetic models (Figures 1 and 2). The curve-fitting results for the high-valent model 3 were quite different from those of the other iron complexes. In this case, a reasonable fit was obtained with only a single shell of nitrogen, with a relatively short (1.92 A) Fe-N. Attempts to improve this fit by including a shell of oxygen at a short distance resulted in the disappearance ( u 1) of either the N or the 0 shell. These fitting results, together with X-ray edge spectra indicating a highly symmetric Fe environment: are most consistent with a symmetric 6-coordinate structure having an average Fe-ligand distance of 1.92 A. The Fe-N(porphyrin) distance expected for a 6-coordinate porphyrin is 1.98 A.56In order to obtain the observed average Fe-ligand distance, this would require an Fe-Oaxlaidistance of

-

--

-

+

(56) Scheidt, W.R.; Reed, C. A. Chem. Rep. 1981, 81, 543-555.

7824 J . Am. Chem. Soc., Vol. 108, No. 24, 1986

Penner-Hahn et ai.

-

1.80 A in 3. This interpretation is supported by recent N M R and Mossbauer data showing that 3 is in fact the bismethoxy adduct of Fe1V(TMP).30BAnalogous dialkoxy complexes of Cr5’ and M d 8 porphyrins have average bond lengths of M-N = 2.01 A and M - O = 1.84 A, for an overall average M-ligand distance of 1.95 A. These distances are consistent with the EXAFS results for 3. The EXAFS study of H R P reported by Chance et aL3’ included several fits in which shells separated by less than 0.1 8, were resolved (for example, HRP-I1 was reported to have Fe-N bond lengths of 2.10 and 2.00 A and an Fe-O bond length of 1.93 A). In our hands, attempts to fit multiple shells of the same atomic type (C, N, or 0) suffer from severe parameter correlation. Nevertheless, we have attempted a similar fitting procedure for the data of 3, in which we used two shells of N/O ligands a t the appropriate distances (1.8 and 2.0 A). Although it was possible to fit the data for 3 with shells located at -1.85 and 1.98 A, the improvement in the mean-square deviation for this fit relative to the single-shell fit was negligible (